Read more.Taylor argues that CPU doesn't do AMD's hardware justice.
Read more.Taylor argues that CPU doesn't do AMD's hardware justice.
Isn't this what AMD has been saying about APU's for years? Doesn't sound anything new to me.
What I want to see is decent performance with the APU running in tandem with a discrete graphics card. I get the feeling that at the moment the GPU part of the APU is just dead silicon wasting energy in that configuration.
If I want something that will run low-res, I have a tablet, and DLNA enabled NAS and a PS3 takes care of the media streaming. Horses for courses I guess.
Corsair Air 540, Asus Prime X570-Pro, Win 10 Pro, AMD R9 3900X, Corsair HX 750, EVGA 1080 Ti, 2x Corsair 2TB MP600, 2x 2TB WD20EZRX, 4x8GB Corsair Dominator, custom watercooled (single loop, 2 rads)
Corsair 550D, Asus X470-Prime Pro, Win 10 Pro, AMD R7 2700, Corsair RM750i, Asus GTX780 Poseidon, 2x Sammy 500GB 970 EVO, 2x 2TB Seagate Barracuda, 2x8GB Corsair Vengeance, custom watercooled (single loop, 2 rads)
Synology DS918+ w/ 2xWDC Green 3TB + 2x Seagate Barracuda 6TB, N2200 w/ 2xSammy 1.5TB
backup:
Corsair 500R, Gigabyte GA-Z97MX Gaming 5, Win 10 Pro, i5 4690, Corsair HX750i, Sapphire Fury X, 256GB Sammy SM951 M.2 (System), WD SE16 640GB, 2x8GB Corsair Vengeance, Corsair H100i
APUs are a solution searching for a problem.
It's a marketing piece. It wouldn't be out of place on the side of the APU box. Not a complaint in any form though, AMD need to get the message out as a lot of people don't really appreciate their APUs.
I wouldn't say that - AMDs APUs are well suited to most home machines as they offer decent graphics at a pretty low price point. The power savings of some of them against a separate CPU and GPU are often also worth it.
I used that A6 3670K I won last year in a family members home PC,and it can run some casual games like Torchlight or Minecraft pretty well. That was a CPU which cost between £60 to £70 and I used bog standard 1600MHZ DDR3 too. Plus when you consider pre-built boxes most have IGPs,it means that even a basic PC nowadays can run far more games than say a few years ago. They are great for low cost general purpose family PCs and I know a lot of people who have done such builds(or bought them from retailers),and they seem very good for the price.
Utterly don't agree with you there - APU's are ideal for those folks who need power efficiency in the desktops and aren't necessarily looking for something that'll run the latest CoD at 60 frames+ at 1080p. Remember that the big deal with these is that you get cpu+gpu in a single module - so that means cost savings for the folks making that desktop you're going to buy.
AMD's APU's seem - to me at least - to be the best of the bunch since they've got a pretty reasonable GPU section. Okay, I'm willing to admit that the Intel equivalents probably are better "compute engines", but they're also more expensive. So you can nip out and buy a desktop based on one of these and know that you'll at least be able to do some gaming without running out to buy an expensive graphics card if you need a break from the office/web/mail uses.
As I said, they won't appeal to anyone who interested in serious PC gaming or outright speed. But then again, Intel's got that segment of the cpu market anyway.
Agree with AMD's push for OpenCL - this strikes me as many times more useful than either DirectCompute or CUDA. "Khronos" though - wasn't that the evil guy's project in "The Incredibles"?
Actually this is new as current APUs are severely hampered by current memory tech. If you look at the PS4/XB1 architecture which in the case of PS4 also has UMA then you have an APU that is at a whole new performance level compared to current PC APU offerings. This gives you the advantages of a CPU & a powerful discrete GPU in a single processor. Once programmers really start to take advantage of this and think of the CPU/GPU as a single processor capable of both complex serial and highly parallel tasks, utilising whichever is the most appropriate based on the type of load we'll see a whole new level of performance. At present this is impossible because of the segregation between CPU & GPU, there being no consistency on the GPU side (everybody has a different one), and their different memory architectures (DDR3 which really bottlenecks APUs). As such CPUs have to handle all general compute tasks including parallel stuff more suited to a GPU, and the GPU is mostly reserved only for graphics. Hopefully the next gen consoles will help speed the transition up since they are both based on newer more advanced AMD APUs.
We did the same a while back for a family member, but IIRC we put some 2133 RAM in as there was quite an increase with the GPU. The RAM didn't really cost any more either, but some of timings were a bit more loose (not real hit though).
I think it was either a 5700 or 5800, but I remember us installing a few games off Steam and having no issues. It helped that his monitor was 720p and didn't see the need to upgrade, but most of the games were perfectly playable.
If you're willing to lower the settings of most games a little, you can kick out some quite nice high FPS.
The power draw was also very good. Once we setup all the standby options / turning the screen off and so on, you'd probably see a noticeable decrease in electricity costs over the period he will keep it.
Crazy when you think about it - some of these APU systems can pull power in the region of a single lightbulb from past generations.
Actually it is hinted at in the article "The APU is a device designed to offer both serial and parallel programming architecture and it’s clearly here to stay. It promises an exciting future, especially with the introduction of new unified memory architectures due next year."
Also I wasn't just referring to UMA. Current APUs use slow DDR3 which bottlenecks the GPU side, PS4 uses DDR5 removing the bottleneck. We may see DDR4/5 becoming mainstream for system RAM next year which will greatly speed up APUs. There are many articles around showing 3D game performance improvements of up to 30% with current AMD APUs by OC'ing RAM from 1600 to 2133.
I agree I have a custom built PC for gaming on but am looking to build another (3 is never enough lol) desktop to use in the sitting room for general use, so this sounds like something I would use for that as it would be a main pc for everything other then PC gaming because as you said Intel has that covered and I don't think I would go back to AMD for gaming.
Don't get me wrong, I use to have an AMD PC for gaming a few years back, but I have found Intel to be better for that use and as a big PC gamer I am happy to pay more for Intel then AMD as I know the extra cash is well worth it for my gaming needs.
I look forward to seeing what AMD offer with their APU's in the next few months when I look around for building that 4th PC for more general use, the odd game and well basically a cheaper to run PC which can also handle lower games such as CS.
OpenCL is at the very core of their argument though. To make their approach valid we need to have a massive rewrite of a load of software to support parallel processing. Apart from the 'intensive' tests that hardware sites love to throw at us when reviewing CPUs, this kind of approach simply isn't going to take off in the consumer sphere. And at the pro sphere, Nvidia simply have better offerings with discrete cards that can be shared between virtual servers.
AMD are bringing a 5 legged donkey from the rescue shelter to the Grand National. It may have a lot of features and be cheap, but the result is an incoherent mess. Just look at their hybrid crossfire.
I disagree with this. Going forward, it makes sense to incorporate OpenCL if possible to optimise programs since many computers have untapped potential in this regard. Having programs utilise the vast resources available opens the door to lower power consumption than having x86 cores chomp through everything and with better performance on the table, I would certainly not complain.
The APU is a good concept, but now AMD really needs to get on the ball and increase both performance of the CPU portion AND the GPU portion of their APUs asap. I say this in light of the fact the latest mobile core i7 haswell CPU with HD5200 integrated graphics absolutely demolishes even a desktop a10 5800k in games. Granted that is an i7 quad in a more expensive ultra book kind of chassis, but if intel can move that level of gpu performance down to the i5 and i3 markets, AMD will be in serious trouble in the sector they have been profiting the most: the lower end budget units. That intel hd5200 can play modern games on at least medium at 720p, The a10 5800k cannot. I know the newest high end desktop APU graphics can outperform the likes of hd6670(ddr3), but that simply isn't enough. I look forward to what both companies have in store in the months and years ahead.
The Adobe CS suite now has OpenCL acceleration as standard now in both Windows and OS X versions and more and more software is starting to adopt it,as it will run on any compliant GPU. OpenCL development was actually started by the tiny tech company known as Apple and has the backing of not only AMD but also small companies like IBM, Qualcomm, Intel, and Nvidia.
OpenCL is already starting to be relevant for more and more consumer software unfortunately for some.
However,for everyone else it is good news.
Anyway,longterm AMD and its partners are looking to develop better tools which would make it easier to programme across different chip types,hence probably bypassing OpenCL,CUDA,etc in the long term.
Also,the HSA foundation has an impressive list of members too:
http://hsafoundation.com/
http://seekingalpha.com/article/1355...erates-for-amd
I saw this on HPCwire recently:
http://www.hpcwire.com/hpcwire/2013-...u_prowess.html
You do realise the HD5200 uses a GPU section larger than than the GPU in an HD7790, made on a expensive 22NM process, with another 84MM2 of L4 cache made on an even more expensive subset of that process?? Its expensive because it is a very big chip by Intel standards made on an expensive process. You are talking about a total die area near that of the chip used in the HD7970(bigger than a GK104 too) made on a cutting edge process.
Intel sells it at such a high price,otherwise it would crater its own margins which need to be high. Why? It needs to spend billions on process node tech which needs to be amortised over every chip they sell,and they rely on it to compete with everyone from AMD to the ARM and MIPs licensees.
The fastest APU for gaming are the XBox One and PS4 SOCs,but these are also quite big too,but they are made on a cheaper TSMC process. TSMC sells more chips than the next three competitors combined(including Intel and GF) and they do so at good margins and massive volume,which spreads their costs out(they spend less on process node R and D than Intel too).
The A10 CPUs are made to hit a certain price point anyway,whereas the Core i7 CPUs with the HD5200 are made to be sold at a much higher price.
Last edited by CAT-THE-FIFTH; 30-08-2013 at 10:55 PM.
There are currently 1 users browsing this thread. (0 members and 1 guests)